Journal of Biogeography (J. Biogeogr.) (2016)

ORIGINAL Little evidence for Bergmann’s rule body ARTICLE size clines in passerines along tropical elevational gradients Benjamin G. Freeman1,2*

1Department of Ecology and Evolutionary ABSTRACT Biology, Cornell University, W257 Corson Aim To test whether intra- and interspecific patterns in body mass along Hall, Ithaca, NY, USA, 2Cornell Lab of elevational gradients follow Bergmann’s rule for a subset of tropical montane Ornithology, 159 Sapsucker Woods Rd, Ithaca, NY, USA passerines. Location Tropical elevational gradients in New Guinea, Borneo, Peru and Costa Rica. Methods I used linear regressions to assess intraspecific patterns in body mass along elevational gradients in common New Guinean passerines (2697 mist- netted individuals of 21 species). I then evaluated interspecific patterns using two data sets. First, I investigated differences in body mass in species pairs of elevational replacements, closely related species with minimal overlap along elevational gradients (species pairs; New Guinea: n = 45, Borneo: n = 22, Peru: n = 58 and Costa Rica: n = 30). Second, I used a comparative phylogenetic approach to test whether species’ mid-point elevations predicted their masses within entire passerine avifaunas found along single elevational gradients (species; New Guinea: n = 184, Peru: n = 529 and Costa Rica: n = 220).

Results New Guinean passerines exhibited minimal intraspecific variation in mass along elevational gradients. In two species, lower elevation individuals had significantly larger masses than upper elevation conspecifics. In species pairs of elevational replacements, there was no trend for upper elevation species to have larger masses than lower elevation species. Overall, species pairs tended to have positive mass disparities (mass of upper elevation species/mass of lower elevation species). However, contrary to predictions of Bergmann’s rule, mass disparity was unrelated to elevational overlap. When considering entire passerine avifaunas along single elevational gradients, species’ masses were uncorrelated with their mid-point elevational distributions. Main conclusions I found little evidence that tropical montane passerines have larger body masses at higher elevations where temperatures are colder. This lack of pattern was consistent across evolutionarily independent avifaunas of different biogeographical regions. These results suggest mean temperature is not a gener- ally important driver of body size evolution in tropical montane passerines. *Correspondence: Benjamin G. Freeman, Department of Ecology and Evolutionary Keywords Biology, Cornell University, W257 Corson Bergmann’s rule, body size, body size clines, comparative phylogenetics, Hall, Ithaca, NY, USA. E-mail: [email protected] ecogeographic rule, elevational gradient, tropical mountains

drive the evolution of body size (Brown et al., 2004), tem- INTRODUCTION perature is one potentially important abiotic factor influenc- Body size is an ecologically influential trait that varies widely ing body size evolution – Bergmann’s rule describes the within and among species (LaBarbera, 1989; Brown et al., pattern that populations or species of endotherms living in 2004). Although many abiotic and biotic mechanisms can colder environments tend to be larger than related

ª 2016 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/jbi 1 doi:10.1111/jbi.12812 B. G. Freeman populations or species living in warmer environments elevations as predicted by Bergmann’s rule. I addressed this (Bergmann, 1847; James, 1970). The proper formulation of question by testing how body sizes in the largest group of Bergmann’s rule remains debated; perhaps most importantly, tropical montane birds – the passerines – are related to their Bergmann’s rule has been considered both a pattern (a nega- elevational distributions. To assess the generality of any pat- tive relationship between body size and temperature) and a terns (or lack thereof), I investigate the relationship between process (temperature exerts selection on body size via physio- body mass and elevational distribution in interspecific com- logical mechanisms such as thermoregulation; e.g. Watt et al., parisons within the evolutionarily distinct avifaunas of the 2010; Meiri, 2011; Olalla-Tarraga, 2011; Watt & Salewski, Neotropics, Southeast Asia and Melanesia. I consider Berg- 2011). Investigations of Bergmann’s rule have traditionally mann’s rule to simply be a negative relationship between analysed body size clines along latitudinal gradients (e.g. body size and temperature (i.e. a pattern; hereafter ‘Berg- Ashton, 2002; Ashton & Feldman, 2003; Watt et al., 2010; mann’s rule’). As such, Bergmann’s rule predicts that (1) Feldman & Meiri, 2014). However, temperature declines not within species, individuals at high elevations should tend to only with increasing latitude, but also with increasing eleva- be larger than individuals at low elevations, (2) within spe- tion; thus, studies have also analysed whether Bergmann’s cies pairs of closely related species, upper elevation species rule patterns are found in body size clines along elevational should be larger than lower elevation species, with this rela- gradients (e.g. Brehm & Fiedler, 2004; Herzog et al., 2013). tionship strongest in species pairs that inhabit non-overlap- Tropical elevational gradients offer an excellent geographi- ping elevational distributions (and thus experience more cal arena to investigate whether body size clines are associ- different ambient temperatures relative to species pairs with ated with temperature. Temperatures decline over short greater elevational overlap) and (3) when accounting for distances along tropical mountain slopes, where ambient phylogenetic relationships, elevational distributions should be mean temperature drops c. 5–6 °C per 1000 m gain in eleva- significantly positively related to body size in large assem- tion (Forero-Medina et al., 2011; Freeman & Class Freeman, blages of species. I tested these predictions using (1) field 2014a). As a consequence, sites on steep slopes may be data for common species of New Guinean understorey located just a few kilometres apart but experience very differ- passerines captured along two single elevational gradients, ent temperatures. Because temperature variation is typically (2) relative body masses in species pairs of closely related minimal at particular sites along tropical elevational gradi- species that inhabit minimally overlapping elevational distri- ents (e.g. daily temperatures at a given site within the forest butions along an elevational gradient in four distinct tropical understorey vary by c. 5 °C, and annual variation is typically montane regions (the Eastern highlands of New Guinea, minor), sedentary tropical organisms separated by even small Manu National Park in the Peruvian Andes, the Caribbean (c. 750 m) expanses of elevation can experience completely slope of Costa Rica and the highlands of Malaysian Borneo) distinct temperatures, at least in the shaded forest interior and (3) phylogenetic comparative methods to assess whether (Janzen, 1967). species’ mid-point elevational distributions significantly pre- Evidence that tropical and subtropical montane faunas dicts their body size in the entire passerine avifaunas found exhibit body size clines consistent with Bergmann’s rule in three regions (the Eastern highlands of New Guinea, along elevational gradients is mixed. Increases in body size in Manu National Park in the Peruvian Andes and the Carib- colder high elevation environments within species have been bean slope of Costa Rica). Taken together, these analyses found in some cases (e.g. Rand, 1936; Lu et al., 2006; Van- provide a general test of whether tropical montane passerines derwerf, 2012), and in interspecific comparisons in clades of conform to Bergmann’s rule. Neotropical frogs (Gouveia et al., 2013) and lizards (Cruz et al., 2005; Zamora-Camacho et al., 2014), but not in clades MATERIALS AND METHODS of Asian frogs (Hu et al., 2011) or Neotropical butterflies (Hawkins & Devries, 1996), moths (Brehm & Fiedler, 2004) Intraspecific or dung beetles (Herzog et al., 2013). Results can be incon- sistent within a taxonomic group in a single geographical Bergmann’s rule predicts that, within species, individuals region. For example, patterns of intraspecific body size varia- should tend to have larger masses at high elevations. I tested tion in Andean birds follow Bergmann’s rule in some (Tray- this prediction using field body mass data gathered along lor, 1950; Graves, 1991; Bulgarella et al., 2007) but not all two elevational gradients in Papua New Guinea: the YUS (Remsen, 1984, 1993) species, including an example where a Conservation Area, Morobe Province and the north-west species exhibits Bergmann’s rule body size clines across lati- ridge of Mt Karimui, Chimbu Province. The YUS Conserva- tudinal but not elevational gradients (Gutierrez-Pinto et al., tion Area (hereafter YUS, approximate coordinates: À6.00, 2014). This inconsistency holds for analyses at the interspeci- 146.84) is located on the northern scarp of the Saruwaged fic level – some clades of tropical birds exhibit Bergmann’s Range on the Huon Peninsula. Between 2010 and 2012, a rule body size clines while others do not (Blackburn & Rug- team of fieldworkers conducted mist-net surveys in primary giero, 2001). forest along a single elevational gradient from 230 to 2940 m Thus, it remains unclear whether tropical montane birds in YUS; a total of 18 mist-net surveys were completed along generally exhibit larger body masses at higher (colder) 1-km trails cut along elevational contours at intervals of

2 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Tropical passerines are not larger at high elevations

120–200 m in elevation (see Freeman et al., 2013, for more used single references to compile species pairs of passerine information). On the north-west ridge of Mt. Karimui (ap- elevational replacements found along forested elevational gra- proximate coordinates: À6.56, 144.76), a team of fieldwork- dients in four regions that feature large mountain ranges ers surveyed the understorey bird communities with mist spanning from lowlands (< 400 m) to above tree line nets along a continuous gradient of primary forest between (> 3000 m): the Eastern highlands of New Guinea (n = 45 1150 m and 2520 in June–July 2012 (see Freeman & Class species pairs; Pratt & Beehler, 2014), Manu National Park in Freeman, 2014b, for further details). At both sites, captured the Andes of south-eastern Peru (n = 58 species pairs; individuals were weighed using Pesola spring scales (a 30-g Walker et al., 2006), the Caribbean slope of Costa Rica scale for smaller species and a 100-g scale for larger species). (n = 30 species pairs; Stiles & Skutch, 1989) and Malaysian I compiled field body mass data for 2548 individuals of 21 Borneo (n = 22; Myers, 2009). I quantified body masses for common understorey passerines from seven families found each species using a single reference volume (Dunning, 2007; within YUS (see Appendix S1 in Supporting Information. see Appendix S2). I limited my analysis to forest passerines These species were commonly captured (mean number of (hereafter ‘passerines’), as the majority of habitat in these weighed individuals = 122, range = 48–327 weighed individ- regions is forested. uals/species) across a range of elevations (mean elevational I first used a sign test to ask whether cases where upper breadth per species = 1220 m, range = 520–1810 m). I addi- elevation species had larger masses than their lower elevation tionally included field body mass data for 139 individuals relatives (‘high and heavy’ species pairs) outnumbered from two species from Mt. Karimui (n = 102 and 37 reversed cases (‘low and large’ species pairs) in each region. weighed individuals; elevational breadth = 800 and 1025 m, This simple analysis does not consider quantitative differ- respectively; these same two species also appear in the YUS ences in mass difference. I therefore calculated the mass dis- data, see Appendix S1). I then used linear regressions imple- parity for each species pair – the mass of the upper elevation mented in R (R Core Team, 2014) to test if, for each species species divided by the mass of the lower elevation species, at each site, individuals captured at higher elevations tended such that positive mass disparities indicate ‘high and heavy’ to be heavier than individuals captured at lower elevations. I cases – and used t-tests to assess whether species pairs within included sex as a predictor variable for four species in which regions had significantly positive mass disparities. This analy- males and females differ in plumage [black berrypecker (Mel- sis, in turn, overlooks differences in the degree to which spe- anocharis nigra), fan-tailed berrypecker (Melanocharis ver- cies within species pairs experience different ambient steri), regent whistler (Pachycephala schlegelii) and black temperatures. If colder temperatures are associated with lar- fantail (Rhipidura atra)], though note that juvenile males ger body masses, as predicted by Bergmann’s rule, body mass have female-like plumage in these species, and applied a disparities should be negatively correlated with elevational Bonferroni correction to account for the influence of multi- overlap (greater mass disparities in species pairs that occupy ple tests on statistical significance. distinct elevational zones that do not overlap). I tested this prediction using a t-test to examine whether species pairs with non-overlapping elevational distributions (n = 63, a Interspecific: Elevational replacements subset of the total data set) had mass disparities significantly Bergmann’s rule predicts that, when closely related species greater than zero, and also a linear regression model, with inhabit different elevational zones, (1) the upper elevation mass disparity as the response variable and elevational over- species should have a larger mass and (2) this difference in lap and region as predictor variables. mass should be positively correlated with elevational diver- gence within the species pair (i.e. upper elevation species Interspecific: Passerine avifaunas should have especially larger masses than their lower eleva- tion relatives when species within a species pair inhabit com- Bergmann’s rule predicts that, in large assemblages of species, pletely distinct elevational zones and thus experience a species’ elevational distributions should be significantly greater difference in temperatures than species pairs with related to their body mass when taking phylogenetic relation- greater elevational overlap). I tested these predictions by ships into account. I tested this prediction using the passer- identifying species pairs of closely related species (nearly all ine avifaunas found along elevational gradients in three congeners, see Appendix S2) that occupied divergent eleva- distinct regions, the Eastern highlands of New Guinea, Manu tional distributions (defined as species pairs with elevational National Park in south-eastern Peru and the Caribbean slope overlap < 50%; most species pairs in this data set had nar- of Costa Rica. I used the same reference volume to quantify row elevational overlaps, with the median elevation over- body masses for each species in each region (Dunning, 2007) lap = 8.8% and the 75th percentile of elevational and single sources (New Guinea: Pratt & Beehler, 2014; overlap = 23.2%). Such species pairs of ‘elevational replace- Costa Rica: Stiles & Skutch, 1989; Peru: Walker et al., 2006) ments’ are prominent in tropical montane faunas (Patterson to define passerine species’ elevational ranges within regions et al., 1998; Pyrcz & Wojtusiak, 2002; Pasch et al., 2013) and (New Guinea: n = 184 species; Peru: n = 529 species; are especially common in birds (Terborgh & Weske, 1975; Costa Rica: n = 220 species), and used elevational mid-point Jankowski et al., 2010; Freeman & Class Freeman, 2014b). I to characterize species’ elevational distributions (see

Journal of Biogeography 3 ª 2016 John Wiley & Sons Ltd B. G. Freeman

Appendix S2). Because body mass data were unavailable for frequent than ‘low and large’ cases (where the lower eleva- many Bornean species, I did not include the passerine avi- tion species had a larger mass; P-values from sign fauna of Malaysian Borneo in this analysis. tests = 0.20–1, Table 1). However, the difference in masses I then used comparative phylogenetic methods to test within a species pair tended to be greater in ‘high and heavy’ whether species’ elevational mid-points were significantly cases compared to ‘low and large’ cases (Fig. 2a) – mass dis- related to their mass while accounting for evolutionary related- parities were significantly positive in Peru (95% confidence ness among species. I transformed both response and predictor interval for mass disparity = 0.0057–0.16; t = 2.16, d.f. = 57, variables using log transformations so that residuals con- P = 0.035) and positive in each of the other three regions formed to a normal distribution. I used a phylogenetic tree (New Guinea: 95% confidence interval for mass dispar- from Jetz et al. (2012) that consisted of passerine taxa with ity = À0.037 to 0.17; d.f. = 44, t = 1.28, P = 0.21; Costa genetic information (‘Hackett sequenced species’), in combi- Rica: 95% confidence interval for mass disparity = À0.015 to nation with phylogenetic generalized least squares (PGLS, 0.17; d.f. = 29, t = 1.70, P = 0.099; Borneo: 95% confidence Martins & Hansen, 1997), implemented using the packages interval for mass disparity = À0.075 to 0.27; d.f. = 15, ‘nlme’ (Pinheiro et al., 2013) and ‘ape’ (Paradis et al., 2004) in the R programming environment (R Core Team, 2014). I 0.004 scaled internal branch lengths according to Pagel’s k model, which estimates the amount of phylogenetic signal present in Species with 0.002 larger mass the evolutionary history of a given character (Pagel, 1999; at higher elevations Blomberg et al., 2003). In this model, the k parameter varies from 0 (no phylogenetic signal or a star phylogeny) to 1 (phy- 0.000 logenetic signal equal to Brownian motion or branch lengths remain unchanged) and therefore provides a convenient mea- -0.002 Species with smaller mass * sure of evolutionary lability for the trait in question. I exam- Regression coefficient at higher elevations * ined residual plots by eye and removed one outlier from the -0.004 analysis of Costa Rican passerines. Results were very similar with and without this outlier; I present results of the model with the outlier excised. I also investigated using a Ornstein– -0.006

Uhlenbeck (OU) model of trait evolution to investigate body 0 500 1000 1500 2000 2500 mass evolution in each region, but OU models failed to con- Elevational midpoint (m) verge and were thus unable to be parameterized (Ho & Ane, 2014). I therefore report only results of Pagel’s k models. Figure 1 Intraspecific elevational patterns in body mass in 21 species of New Guinean passerines found along two elevational gradients. Each species’ coefficient and 95% confidence interval RESULTS from a mass elevation regression is plotted; points above the zero line indicate species that tended to have larger masses at Intraspecific higher elevations. Asterisks denote two species with significant intraspecific body size clines (P < 0.05) following Bonferroni Most species exhibited minimal variation in body mass along correction (see Fig. S1). New Guinean elevational gradients. Coefficients for species’ mass elevation regressions are clustered around 0, and 95% confidence intervals overlap with 0 in nearly all cases, indi- Table 1 Regional patterns of body mass variation in elevational cating that body size clines along elevation were rare in this replacements of tropical montane passerines. Species pairs were sample (Fig. 1, see Table S1 in Appendix S3 for full results). classified as ‘high and heavy’ when the upper elevation species After using a Bonferroni correction to account for multiple had a larger mass and ‘low and large’ when the lower elevation tests, only two species (out of 23) showed significant eleva- species had a larger mass. P-values give results from sign tests within regions. Two species pairs from Peru had identical body tional body size clines. Both cases were opposite to that pre- masses and body masses were unavailable for six species pairs dicted by Bergmann’s rule; the little shrikethrush from Borneo. (Colluricincla megarhyncha), found in the lowlands and foot- hills, and the rufous-backed honeyeater (Ptiloprora guisei) ‘High and heavy’ ‘Low and large’ found at middle and upper elevations, are each c. 12% smal- Region species pairs species pairs P-value ler in mass at their high elevation limits in YUS compared to New Guinea 23 22 1 their low elevation limits (see Figure S1 in Appendix S3). (Eastern Highlands) Peru (Manu 28 28 1 National Park) Interspecific Costa Rica 19 11 0.20 (Caribbean slope) In each region, ‘high and heavy’ cases (where the upper ele- Borneo (Sabah) 6 10 0.45 vation species had a larger mass) were not significantly more

4 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Tropical passerines are not larger at high elevations

(a) (b) Peru * New Guinea Costa Rica 1.0 1.0 Borneo

0.5 0.5 Mass disparity Mass disparity 0.0 0.0

-0.5 -0.5

Borneo Costa Rica New Guinea Peru 0.0 0.1 0.2 0.3 0.4 n=22 n=30 n=45 n=58 Elevational overlap

Figure 2 Interspecific elevational patterns in body mass in species pairs of elevational replacements found in four regions: (a) boxplots illustrate mass disparities for species pairs in each region. The median is denoted with a horizontal black bar, boxes demarcate first and third quartiles and points and dotted lines illustrate minimum and maximum values. Median values are close to zero, indicating similar numbers of cases with positive (upper elevation species has larger mass) and negative mass disparities (lower elevation species has larger mass), but mean mass disparity is positive in each region, and significantly so in Peru (P < 0.05, denoted by an asterisk). (b) Elevational overlap is unrelated to mass disparity within species pairs of elevational replacements in four regions, indicating that species pairs that experience more divergent temperatures do not tend to have greater mass disparities. The dashed horizontal line at mass disparity = 0 serves to distinguish species pairs with positive and negative mass disparities. t = 1.21, P = 0.24). However, mass disparities were not lar- from zero, indicating that log mid-point elevation does not ger in species pairs with less elevational overlap (Fig. 2b). predict log body mass at the large phylogenetic scale of entire When considering the subset of species pairs with non-over- passerine avifaunas found along an elevational gradient. lapping elevational distributions, mass disparities were not significantly different from zero (n = 63 species pairs; 95% DISCUSSION confidence interval = À0.039 to 0.089, t = 0.79, d.f. = 61, P = 0.43, see points at elevational overlap = 0 in Fig. 2b). In Tropical montane passerines vary widely in body mass (Dun- addition, parameter estimates for elevational overlap were ning, 2007). However, I found little evidence that this varia- not significantly different from zero in a linear regression tion in body mass is related to species’ elevational model predicting mass disparity (Fig. 2b, Table S2 in distributions, which serve as a convenient proxy for the Appendix S3). ambient temperatures experienced by a species. At the Phylogenetic generalized least square models for each intraspecific level, a sample of New Guinean passerines region had lambda values very near 1, indicating relatively exhibited minimal variation in body mass along single eleva- high phylogenetic signal in passerine body mass (Fig. 3, tional gradients (Fig. 1), with two case examples wherein Table S3 in Appendix S3). Parameter estimates for log mid- species showed significant decreases in body mass with eleva- point elevation in each model were not significantly different tion. At the interspecific level, species’ body masses were

6 Peru New Guinea Costa Rica 5

4

Log mass (g) 3

Figure 3 Log elevational mid-point is unrelated to log mass in the passerine 2 avifaunas of three regions. Parameter estimates for elevational mid-point in phylogenetic generalized least squares 0 500 1000 1500 2000 2500 3000 3500 models were not significantly different from zero (see Table S3). Elevational midpoint (m)

Journal of Biogeography 5 ª 2016 John Wiley & Sons Ltd B. G. Freeman weakly related to their elevational distributions at a shallow to have a greater difference in mass (mass disparity), and this phylogenetic scale (when considering species pairs of eleva- relationship was significantly positive in one region (Peru). tional replacements, closely related species pairs that inhabit These results provide some support for a positive relationship different elevational zones; Table 1, Fig. 2), and unrelated to between body size and elevation. However, the key prediction their body masses at deeper phylogenetic scales (considering of the Bergmann’s rule pattern, applied to species pairs of ele- entire passerine avifaunas; Fig. 3). Given that the tropical vational replacements, is that mass disparities are largest in montane passerines in this study did not demonstrate body species pairs that experience very different temperatures (i.e. size clines consistent with Bergmann’s rule, the mechanistic species pairs that inhabit non-overlapping elevational distribu- processes of physiological adaptation hypothesized to under- tions). I found no evidence that this was the case in any region lie Bergmann’s rule seem unlikely to generally apply to tropi- (Fig. 2b). Additionally, species’ elevational mid-points were cal montane birds. not related to their body mass in a comparative phylogenetic These results contrast with previous published studies analysis of entire passerine avifaunas (Fig. 3). investigating body mass variation in birds that have often On the surface, these results contradict a previous study of found geographical patterns of body size clines consistent Andean passerines that found correlations between species’ with Bergmann’s rule. For example, global analyses of body masses and elevational mid-points (Blackburn & Rug- intraspecific variation in avian body mass have found the giero, 2001). These contradictory results could be due to dif- strong pattern that populations in colder environments are ferences in spatial scale between analyses – instead of the typically larger than those found in warmer environments entire avifauna found within a region (i.e. all Andean passeri- (Ashton, 2002; Meiri & Dayan, 2003). However, these studies nes), I used only the set of species found along a single eleva- included few tropical species and primarily considered body tional gradient in south-eastern Peru in my analysis. mass patterns along latitudinal gradients, where differences in However, in the regional analysis, mid-point elevation temperature covary with many additional abiotic (e.g. tem- explained only 2% of variation in body mass, and many clades perature seasonality) and biotic (e.g. species richness, resource did not follow Bergmann’s rule (Blackburn & Ruggiero, seasonality) factors that could also influence body size evolu- 2001). Thus, Blackburn & Ruggiero’s (2001) analysis, while tion. Intraspecific patterns in tropical birds along elevational supporting the existence of a weak Bergmann’s rule pattern in gradients are sometimes consistent with Bergmann’s rule (e.g. Andean passerines, also suggests that tropical montane passer- Traylor, 1950; Vanderwerf, 2012), but most New Guinean ines do not consistently show body size clines concordant passerines (21 out of 23 comparisons) in this study did not with Bergmann’s rule. This view accords with previous studies vary in mass over an elevational gradient, and the two excep- that found body size patterns in other tropical montane fau- tions were in the opposite direction to that predicted by Berg- nas do not conform to Bergmann’s rule (Hawkins & Devries, mann’s rule. These results conflict with a previous analysis of 1996; Brehm & Fiedler, 2004; Herzog et al., 2013). New Guinean birds that reported body size increases (using These results have implications for the relationship between wing length as the metric of body size) with elevation in a elevational distribution and competitive dominance in birds, wide variety of species in the mountains of southeast New where behavioural dominance in interspecific contests is typi- Guinea, including several species included in the current anal- cally associated with body size (Robinson & Terborgh, 1995; ysis (Rand, 1936). This discrepancy could reflect different Freshwater et al., 2014; but see Martin & Ghalambor, 2014). patterns of body size variation in different montane regions Recent field experiments have supported the long-standing within New Guinea or different methodologies (measuring hypothesis (Terborgh & Weske, 1975) that asymmetric inter- body mass vs. wing length). However, it seems more likely specific aggression can influence the elevational distributions of that the patterns described by Rand (1936) may be artefacts pairs of tropical elevational replacements (Jankowski et al., of the small sample sizes of museum specimens then available 2010; Pasch et al., 2013). Many tropical montane passerines are for analysis (mean = c. 10–25 individuals/species in Rand’s shifting their distributions upslope associated with recent warm- analysis vs. mean = 122 individuals/species in the current ing (Forero-Medina et al., 2011; Freeman & Class Freeman, analysis). Nevertheless, further studies are necessary to test 2014a), and it has been hypothesized that asymmetric inter- the possibility that intraspecific body size clines consistent specific aggression between tropical elevational replacements with Bergmann’s rule are present in tropical montane passeri- may influence their rates of warming-associated upslope shifts nes in species omitted in my analysis (e.g. canopy species that (Jankowski et al., 2010 Freeman et al. 2016). I found no consis- are poorly sampled with mist nets), in other regions in New tent pattern in relative body mass between upper and lower ele- Guinea, or more generally in other tropical regions. vation species pairs of elevational replacements. Thus, though When comparing closely related species that inhabit differ- speculative, if body mass is associated with behavioural domi- ent elevational distributions within each of four regions, I nance in tropical avian elevational replacements, relative eleva- found equal proportions of cases where the upper elevation tional distribution alone is unlikely to predict interspecific species had a larger mass (‘high and heavy’) and where the aggression in these taxa. It is therefore likely that field studies lower elevation species had a larger mass (‘low and large’; will demonstrate both instances where lower elevation species Table 1). While the proportion of ‘high and heavy’ and ‘low are larger and behaviourally dominant (and could potentially and large’ cases was similar, ‘high and heavy’ examples tended ‘push’ their upper elevation replacement upslope with

6 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Tropical passerines are not larger at high elevations continued warming; e.g. Catharus thrushes in Jankowski et al., in an Andean montane rain forest. Global Ecology and 2010 and songbirds in Freeman et al. in press; also see Freeman Biogeography, 13,7–14. & Montgomery, 2016 for a possible temperate zone example) Brown, J.H., Gillooly, J.F., Allen, A.P., Savage, V.M. & West, and also cases where upper elevation species are larger and G.B. (2004) Toward a metabolic theory of ecology. behaviourally dominant (and may be able to maintain their dis- Ecology, 85, 1771–1789. tributions in the face of continued warming as ‘kings of the Bulgarella, M., Wilson, R.E., Kopuchian, C., Valqui, T.H. & mountain’; e.g. Scotinomys singing mice in Pasch et al.,2013). McCracken, K.G. (2007) Elevational variation in body size In conclusion, I found little evidence that tropical montane of Crested Ducks (Lophonetta specularoides) from the Cen- passerines conform to Bergmann’s rule – in several analyses, tral High Andes, Mendoza, and Patagonia. Ornitologıa the body masses of tropical montane passerines were unrelated Neotropical, 18, 587–602. to the elevational zones they inhabit. I found this lack of pat- Cruz, F.B., Fitzgerald, L.A., Espinoza, R.E. & Schulte, J.A. tern in both intraspecific (New Guinean birds) and interspeci- (2005) The importance of phylogenetic scale in tests of fic analyses (passerine avifaunas in multiple largely Bergmann’s and Rapaport’s rules: lessons from a clade of evolutionarily independent biogeographical regions). Because South American lizards. Journal of Evolutionary Biology, body size clines in tropical montane passerines do not conform 18, 1559–1574. to the pattern predicted by Bergmann’s rule, the hypothesized Dunning, J.B. (2007) CRC handbook of avian body masses, process of colder mean temperatures selecting for larger body 2nd edn. CRC Press, Boca Raton, FL. sizes due to physiological factors is unlikely to generally apply Feldman, A. & Meiri, S. (2014) Australian snakes do not fol- to tropical montane passerines. Thus, mean temperature low Bergmann’s rule. Evolutionary Biology, 41, 327–335. appears to exert a minimal (or idiosyncratic) influence on Forero-Medina, G., Terborgh, J., Socolar, S.J. & Pimm, S.L. body size in tropical montane passerines. In this view, biotic (2011) Elevational ranges of birds on a tropical montane factors (e.g. social selection, resource availability and species gradient lag behind warming temperatures. PLoS ONE, 6, interactions), and the interplay between abiotic and biotic fac- e28535. tors, may be more important drivers of body mass evolution Freeman, B.G. & Class Freeman, A.M. (2014a) Rapid upslope in tropical montane birds. shifts in New Guinean birds illustrate strong distributional responses of tropical montane species to global warming. 111 ACKNOWLEDGEMENTS Proceedings of the National Academy of Sciences USA, , 4490–4494. I thank two anonymous referees for comments that greatly Freeman, B.G. & Class Freeman, A.M. (2014b) The avifauna improved this manuscript, T. Heaton and E. Sibbald for of Mt. Karimui, Chimbu Province, Papua New Guinea, assistance compiling body mass data and N.A. Mason for including evidence for long-term population dynamics in statistical advice. This material is based on work supported undisturbed tropical forest. Bulletin of the British Ornithol- by a National Science Foundation Graduate Research Fellow- ogists’ Club, 134,30–51. ship, Award No. 2011083591 and a National Science Founda- Freeman, B.G. & Montgomery, G. (2016) Interspecific tion Postdoctoral Fellowship in Biology, Award No. 1523695. aggression by Swainson’s Thrush (Catharus ustulatus) may limit the distribution of the threatened Bicknell’s Thrush (Catharus bicknelli) in the Adirondack Mountains. The REFERENCES Condor: Ornithological Applications, 118, 169–178. Ashton, K.G. (2002) Patterns of within-species body size Freeman, B.G., Class Freeman, A.M., & Hochachka, W.M. variation of birds: strong evidence for Bergmann’s rule. (2016). Asymmetric interspecific aggression in New Gui- Global Ecology and Biogeography, 11, 505–523. nean songbirds that replace one another along an eleva- Ashton, K.G. & Feldman, C.R. (2003) Bergmann’s rule in tional gradient. Ibis. DOI: 10.1111/ibi.12384 nonavian reptiles: turtles follow it, lizards and snakes Freeman, B.G., Class, A.M., Mandeville, J., Tomassi, S. & reverse it. Evolution, 57, 1151–1163. Beehler, B.M. (2013) Ornithological survey of the moun- Bergmann, C. (1847) Ueber die verh€altnisse der tains of the Huon Peninsula, Papua New Guinea. Bulletin w€armeokonomie€ der thiere zu ihrer grosse.€ Gottinger Stu- of the British Ornithologists’ Club, 133,4–18. dien, 3, 595–708. Freshwater, C., Ghalambor, C.K. & Martin, P.R. (2014) Blackburn, T.M. & Ruggiero, A. (2001) Latitude, elevation Repeated patterns of trait divergence between closely and body mass variation in Andean passerine birds. Global related dominant and subordinate bird species. Ecology, Ecology and Biogeography, 10, 245–259. 95, 2334–2345. Blomberg, S.P., Garland, T. & Ives, A.R. (2003) Testing for Gouveia, S.F., Dobrovolski, R., Lemes, P., Cassemiro, F.A.S. phylogenetic signal in comparative data: behavioral traits & Diniz-Filho, J.A.F. (2013) Environmental steepness, tol- are more labile. Evolution, 57, 717–745. erance gradient, and ecogeographical rules in glassfrogs Brehm, G. & Fiedler, K. (2004) Bergmann’s rule does not (Anura: Centrolenidae). Biological Journal of the Linnean apply to geometrid moths along an elevational gradient Society, 108, 773–783.

Journal of Biogeography 7 ª 2016 John Wiley & Sons Ltd B. G. Freeman

Graves, G.R. (1991) Bergmann’s rule near the equator: latitu- Olalla-Tarraga, M.A. (2011) ‘Nullius in Bergmann’ or the dinal clines in body size of an Andean passerine bird. Pro- pluralistic approach to ecogeographical rules: a reply to ceedings of the National Academy of Sciences USA, 88, Watt et al. (2010). Oikos, 120, 1441–1444. 2322–2325. Pagel, M. (1999) Inferring the historical patterns of biological Gutierrez-Pinto, N., McCracken, K.G., Alza, L., Tubaro, P., evolution. Nature, 401, 877–884. Kopuchian, C., Astie, A. & Cadena, C.D. (2014) The valid- Paradis, E., Claude, J. & Strimmer, K. (2004) APE: analyses ity of ecogeographical rules is context-dependent: testing of phylogenetics and evolution in R language. Bioinformat- for Bergmann’s and Allen’s rules by latitude and elevation ics, 20, 289–290. in a widespread Andean duck. Biological Journal of the Lin- Pasch, B., Bolker, B.M. & Phelps, S.M. (2013) Interspecific nean Society, 111, 850–862. dominance via vocal interactions mediates altitudinal Hawkins, B.A. & Devries, P.J. (1996) Altitudinal gradients in zonation in Neotropical singing mice. The American Natu- the body sizes of Costa Rican butterflies. Acta Oecologica, ralist, 182, E161–E173. 17, 185–194. Patterson, B.D., Stotz, D.F., Solari, S., Fitzpatrick, J.W. & Herzog, S.K., Hamel-Leigue, A.C., Larsen, T.H., Mann, D.J., Pacheco, V. (1998) Contrasting patterns of elevational Soria-Auza, R.W., Gill, B.D., Edmonds, W.D. & Spector, S. zonation for birds and mammals in the Andes of south- (2013) Elevational distribution and conservation eastern Peru. Journal of Biogeography, 25, 593–607. biogeography of phanaeine dung beetles (Coleoptera: Pinheiro, J., Bates, D., DebRoy, S.S. & Sarkar, D. & R Core Scarabaeinae) in Bolivia. PLoS ONE, 8, e64963. Team (2013) Nlme: linear and nonlinear mixed effects mod- Ho, L.S.T. & Ane, C. (2014) Intrinsic inference difficulties els. Available at http://CRAN.R-project.org/package=nlme for trait evolution with Ornstein-Uhlenbeck models. R package version 3.1-117. Methods in Ecology and Evolution, 5, 1133–1146. Pratt, T.K. & Beehler, B.M. (2014) Birds of New Guinea, 2nd Hu, J., Xie, F., Li, C. & Jiang, J. (2011) Elevational patterns edn. Princeton University Press, Princeton, NJ. of species richness, range and body size for spiny frogs. Pyrcz, T.W. & Wojtusiak, J. (2002) The vertical distribution of PLoS ONE, 6, e19817. pronophiline butterflies (Nymphalidae, Satyrinae) along an James, F.C. (1970) Geographic size variation in birds and it’s elevational transect in Monte Zerpa (Cordillera de Merida, relationship to climate. Ecology, 51, 365–390. Venezuela) with remarks on their diversity and parapatric Jankowski, J.E., Robinson, S.K. & Levey, D.J. (2010) distribution. Global Ecology and Biogeography, 11,211–221. Squeezed at the top: interspecific aggression may constrain R Core Team (2014) R: A language and environment for sta- elevational ranges in tropical birds. Ecology, 91, 1877– tistical computing. R Foundation for Statistical Computing, 1884. Vienna, Austria. Available at: http://www.R-project.org/. Janzen, D.H. (1967) Why mountain passes are higher in the Rand, A.L. (1936) Results of the Archbold Expeditions. No. tropics. The American Naturalist, 101, 233–249. 12. Altitudinal variation in New Guinea birds. American Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K. & Mooers, Museum Novitates, 890,1–14. A.O. (2012) The global diversity of birds in space and Remsen, J.V. (1984) Geographic variation, zoogeography, and time. Nature, 491, 444–448. possible rapid evolution in some Cranioleuca spinetails LaBarbera, M. (1989) Analyzing body size as a factor in ecol- (Furnariidae) of the Andes. Wilson Bulletin, 96, 515–523. ogy and evolution. Annual Review of Ecology and Systemat- Remsen, J.V. (1993) Zoogeography and geographic variation of ics, 20,97–117. Atlapetes rufinucha (Aves, Emberizinae), including a Lu, X., Li, B. & Liang, J.J. (2006) Comparative demography distinctive new subspecies, in southern Peru and Bolivia. Pro- of a temperate anuran, Rana chensinensis, along a relatively ceedings of the Biological Society of Washington, 106,429–435. fine elevational gradient. Canadian Journal of Zoology, 84, Robinson, S.K. & Terborgh, J. (1995) Interspecific aggression 1789–1795. and habitat selection by Amazonian birds. Journal of Ani- Martin, P.R. & Ghalambor, C.K. (2014) When David beats mal Ecology, 64,1–11. Goliath: the advantage of large size in interspecific aggres- Stiles, F.G. & Skutch, A.F. (1989) A guide to the birds of sive contests declines over evolutionary time. PLoS ONE, Costa Rica. Cornell University Press, Ithaca, NY. 9, e108741. Terborgh, J. & Weske, J.S. (1975) Role of competition in dis- Martins, E.P. & Hansen, T.F. (1997) Phylogenies and tribution of Andean birds. Ecology, 56, 562–576. the comparative method: a general approach to incorpo- Traylor, M.A. (1950) Altitudinal variation in Bolivian birds. rating phylogenetic information into the analysis of inter- The Condor, 52, 123–126. specific data. The American Naturalist, 149, 646–667. Vanderwerf, E. (2012) Ecogeographic patterns of morpholog- Meiri, S. (2011) Bergmann’s rule – what’s in a name? Global ical variation in Elepaios (Chasiempis spp.): Bergmann’s, Ecology and Biogeography, 20, 203–207. Allen’s, and Gloger’s rules in a microcosm. Ornithological Meiri, S. & Dayan, T. (2003) On the validity of Bergmann’s Monographs, 73,1–34. rule. Journal of Biogeography, 30, 331–351. Walker, B., Stotz, D.F., Pequeno, T. & Fitzpatrick, J.W. Myers, S. (2009) Birds of Borneo. Princeton University Press, (2006) Birds of the Manu Biosphere Reserve. Fieldiana: Princeton, NJ. Zoology, 110,23–49.

8 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Tropical passerines are not larger at high elevations

Watt, C. & Salewski, V. (2011) Bergmann’s rule encompasses Appendix S2 Database of species pairs of elevational mechanism: a reply to Olalla-Tarraga (2011). Oikos, 120, replacements and all species included in comparative phylo- 1445–1447. genetic analysis. Watt, C., Mitchell, S. & Salewski, V. (2010) Bergmann’s rule; Appendix S3 Additional figures and tables. a concept cluster? Oikos, 119,89–100. Zamora-Camacho, F.J., Reguera, S. & Moreno-Rueda, G. BIOSKETCH (2014) Bergmann’s rule rules body size in an ectotherm: heat conservation in a lizard along a 2200-metre elevational Benjamin Freeman studies the ecological and evolutionary gradient. Journal of Evolutionary Biology, 27, 2820–2828. processes that generate the distributional patterns we observe in modern biotas. His research uses tropical montane avifau- nas as a model system to test theories of biodiversity and SUPPORTING INFORMATION understand species’ responses to environmental change. Additional Supporting Information may be found in the online version of this article: Editor: Walter Jetz Appendix S1 Raw data of New Guinean species’ masses from mist-net captures.

Journal of Biogeography 9 ª 2016 John Wiley & Sons Ltd